雷达资料同化对2015年台风彩虹数值模拟改进

基于WRF中尺度模式，采用集合卡尔曼滤波方法同化中国岸基多普勒天气雷达径向速度资料，对2015年登陆台风彩虹（1522）进行数值试验。从台风强度、路径、结构等方面验证了同化效果，并对不同区域雷达观测资料的同化敏感性进行讨论。试验结果表明:在同化窗内同化分析场台风位置误差相比未同化平均减小15 km，最多时刻减小38 km，同化资料时次越多，确定性预报路径误差越小。同化雷达资料后较好地反映出台风彩虹（1522）近海加强过程，台风中心最低气压同化分析和预报误差相比未同化最大减小超过25 hPa，台风眼的尺度、眼墙处对流非对称结构相比未同化与观测更加接近。试验还表明:台风内核100 km范围内的雷达观测对同化效果影响最大，仅同化这部分资料（约占总量的20%）各方面效果与同化全部资料相近，而仅同化100 km以外资料效果明显不及同化所有资料。仅同化台风内核雷达观测资料可以在不影响同化效果的前提下，使集合同化计算机时减小为原来的1/3，该策略可为台风实际业务预报提供一定参考。

Improving the Simulation of Typhoon Mujigae (2015) Based on Radar Data Assimilation

Typhoon intensity and precise structure are hardly to predict by all kinds of numerical model, and one key problem is the lack of precise initialization data. Through a WRF-based ensemble Kalman filtering (EnKF) data assimilation system, impacts of assimilating China''s coastal Doppler radar velocity observations for track, intensity and structure of Typhoon Mujigae (2015) is examined.Furthermore, assimilating sensitivity of observations in relative regions are also explored. The experimental results show that mean track error and max track error is reduced by 15 km and 38 km, respectively. The track error of the EnKF analysis becomes smaller with more cycles of assimilating data, and so do the deterministic forecast driven by EnKF analysis field. Through data assimilation, offshore enhancement process in Mujigae is well simulated. Intensity error in both EnKF analysis and prediction are smaller than 25 hPa after assimilation. After 9 h cycling radar velocity data assimilation, the deterministic forecast shows the typhoon continue to strengthen before landfall, and the typhoon eye is contracted much after data assimilation. The diameter of typhoon eye is reduced by about 70 km, and the eye wall convection asymmetric structure is closer to observation.The sensitivity of radar observation assimilation is tested by different radial distance area. Numerical sensitivity experiments show that radar observations within 100 km of the typhoon''s inner core play a dominate role to assimilation results. Typhoon track, intensity and structure are all closer to observation by assimilating radar data within 100 km from typhoon center (about 20% of total observation) showing equivalent effects as assimilating all data. Typhoon is somewhat modified by cycling assimilating observations within 100-200 km from typhoon center. There is no obvious enhancement in typhoon track, intensity and structure after assimilating data 200 km away from inner core. Therefore, radar observation located in typhoon kernel is the key to determine assimilation effects. Because of less data assimilated, the strategy of only assimilating inner core radar data can reduce computing time to 1/3 of all data with somewhat same assimilation result. Efficiency of radar assimilation can be much improved by this radar assimilating strategy, and it can give reference to official typhoon real-time data assimilation and prediction work.